基于不完全观测数据的多速率多传感器数据融合

Multirate multisensor data fusion based on missing measurements

研究了一类时变线性动态系统,在不同传感器以不同采样率对同一目标进行观测,并且各个传感器的观测数据存在不规律丢失情况下,给出了一种有效的信息融合方法。该方法通过数学推导,将多速率传感器数据融合转化为单速率传感器数据融合问题,并采用修正的联邦Kalman滤波器进行状态估计。新算法不需要对状态或观测进行扩维,计算量适当,从而保证了算法的实时性。在观测数据丢失的时刻,采用外推的观测值代替错误的观测数据,从而避免了传统算法的发散。理论分析和仿真实验验证了算法的有效性。

A dynamic time-vary linear system is studied.An effective data fusion algorithm is presented in times of multiple sensors observing a single target with different sampling rates.The robustness of the algorithm in case of data missing is also considered,where measurements from each sensor are missing stochastically with certain probabilities.By technical processing,the multirate data fusion system is transformed into a single rate linear dynamic system.By use of the modified federated Kalman filter to the newly established system,the state estimation is obtained.The augmentation of state or measurement dimensions are avoided by use of the presented algorithm,and the real-time property is guaranteed.In times of measurements missing,the observation is replaced by the predicted one,and the divergence of the traditional Kalman filter is omitted.Theoretical analysis and simulation results show the effectiveness of the proposed algorithm.